Re: ice restart ?

See inline

> Emil Ivov <mailto:emcho@jitsi.org>
> April 26, 2014 at 9:40 AM
>
>
> On 26.04.14, 14:42, Robin Raymond wrote:
>> [...]
> Personally I don't see much need for changing ICE credentials on an 
> existing transport. I'd expect an ICE restart to always end up with a 
> new ICE transport.
[RR] This I would much prefer.
>
>> [...]
>
> Well actually ... (and I apologise I am still catching up with ORTC's 
> intended semantics) ... why can't we just use the ICE transport as the 
> object that provides context grouping? Is this what you were 
> suggesting as well?
[RR] Context grouping is not enough. ICE freezing has an ordering. The 
simple implicit ordering I was going to use is based upon the 
RTCIceTransport construction order. Quite simply, if the RTCIceTransport 
within the same context grouping was constructed later it would have 
freezing dependency on those within the same context created earlier.

[RR] But if we force a new RTCIceTransport object to be constructed to 
do ICE restart that new transport will be put at the end of the freezing 
order for the same context. I don't know if that matters since or not 
ultimately (e.g. if we assume that ice restarts would happen on all 
transports within the same context thus they all get created again 
anyway so the order could be preserved by recreating in the objects in 
the same order).
>
> Emil
>
>